Public Abstract

DE-SC0024520: STREAMLINE Collaboration: Machine Learning for Nuclear Many-Body Systems

Award Status: Active

Institution: North Carolina State University, Raleigh, NC
UEI: U3NVH931QJJ3
DUNS: 042092122

Most Recent Award Date: 09/13/2023
Number of Support Periods: 1
PM: Farkhondeh, Manouchehr

Current Budget Period: 09/01/2023 - 08/31/2024
Current Project Period: 09/01/2023 - 08/31/2025
PI: Konig, Sebastian

Supplement Budget Period: N/A

Public Abstract

The goal of the STREAMLINE (SmarT Reduction and Emulation Applying Machine Learning In Nuclear Environments) collaboration is to advance the frontiers of theoretical and computational research on the nuclear many-body problem using machine learning (ML). We have assembled a team that includes leading researchers in innovations and applications of ML to theoretical nuclear physics and one of the leading pioneers in ML algorithms. The scientific problems we address are among the most challenging in computational nuclear many-body theory. The U.S. government has initiated a broad-based, multidisciplinary, multi-agency program to build a sustained national artificial intelligence (AI) structure. In alignment with this initiative, the DOE Office of Nuclear Physics has held several workshops and committees to identify scientific opportunities for AI/ML as well as challenges in applying AI/ML to nuclear physics and high-performance computing. Our collaboration will focus on implementing many of these new research initiatives for nuclear theory. By taking advantage of opportunities offered by ML techniques, STREAMLINE will advance large-scale nuclear physics computations to dramatically increase predictive power and improve our understanding of nuclear structure and dynamics, bulk nucleonic matter, and emergent many-body phenomena. Our ML studies of strongly correlated nucleonic matter will impact experimental programs throughout the U.S. Our team will perform research in the areas of fast and accurate emulators, smart model extrapolation, learning correlations in wave functions and operators, and predicting nuclear dynamics, including nuclear fission, and heavy-ion fusion. The report of the ``AI for Nuclear Physics Workshop'' in 2020 emphasized the need to develop and sustain an AI capable workforce within nuclear physics. To this end, we have planned an extensive educational program for the next generation of nuclear theorists as well as training opportunities for senior researchers. In all of its research and education activities, the STREAMLINE collaboration will devote resources, effort, and training to improve Diversity, Equity, Inclusion, and Belonging in the nuclear physics community and beyond. We aim to serve as a pilot model that builds the intellectual infrastructure and workforce, thereby leveraging future funding initiatives in AI/ML by the DOE Office of Nuclear Physics. The institutions receiving support for the STREAMLINE collaboration are Michigan State University, Argonne National Laboratory, Fermi National Accelerator Laboratory, Florida State University, North Carolina State University, Oak Ridge National Laboratory, Ohio State University, Ohio University, and University of Tennessee.

The goal of the STREAMLINE (SmarT Reduction and Emulation Applying Machine Learning In Nuclear Environments) collaboration is to advance the frontiers of theoretical and computational research on the nuclear many-body problem using machine learning (ML). We have assembled a team that includes leading researchers in innovations and applications of ML to theoretical nuclear physics and one of the leading pioneers in ML algorithms. The scientific problems we address are among the most challenging in computational nuclear many-body theory.

The U.S. government has initiated a broad-based, multidisciplinary, multi-agency program to build a sustained national artificial intelligence (AI) structure. In alignment with this initiative, the DOE Office of Nuclear Physics has held several workshops and committees to identify scientific opportunities for AI/ML as well as challenges in applying AI/ML to nuclear physics and high-performance computing. Our collaboration will focus on implementing many of these new research initiatives for nuclear theory. By taking advantage of opportunities offered by ML techniques, STREAMLINE will advance large-scale nuclear physics computations to dramatically increase predictive power and improve our understanding of nuclear structure and dynamics, bulk nucleonic matter, and emergent many-body phenomena. Our ML studies of strongly correlated nucleonic matter will impact experimental programs throughout the U.S.

Our team will perform research in the areas of fast and accurate emulators, smart model extrapolation, learning correlations in wave functions and operators, and predicting nuclear dynamics, including nuclear fission, and heavy-ion fusion. The report of the ``AI for Nuclear Physics Workshop'' in 2020 emphasized the need to develop and sustain an AI capable workforce within nuclear physics. To this end, we have planned an extensive educational program for the next generation of nuclear theorists as well as training opportunities for senior researchers. In all of its research and education activities, the STREAMLINE collaboration will devote resources, effort, and training to improve Diversity, Equity, Inclusion, and Belonging in the nuclear physics community and beyond. We aim to serve as a pilot model that builds the intellectual infrastructure and workforce, thereby leveraging future funding initiatives in AI/ML by the DOE Office of Nuclear Physics.

The institutions receiving support for the STREAMLINE collaboration are Michigan State University, Argonne National Laboratory, Fermi National Accelerator Laboratory, Florida State University, North Carolina State University, Oak Ridge National Laboratory, Ohio State University, Ohio University, and University of Tennessee.